Statement on US government directive to suspend access to Fable 5 and Mythos 5

Original Article ↗ Hacker News Discussion ↗

Government Order & Anthropic’s Response

  • US government issued an export-control-style directive: Fable 5 / Mythos 5 must not be used by any foreign national, anywhere, including Anthropic’s own foreign employees.
  • Anthropic says the practical effect is shutting the models off for everyone, since they can’t reliably verify citizenship.
  • Government concern appears to be a jailbreaking method that lets the model find software vulnerabilities; Anthropic claims the demo only found simple, already-known issues that other public models can also find without jailbreaks.
  • Legal basis and process are seen as unclear and ad hoc; some compare it to past export controls on cryptography and ITAR.

Reactions to the Government & Politics

  • Many view this as authoritarian overreach, retaliation over earlier Anthropic–Pentagon clashes, or part of a broader anti-immigrant / “America First” stance.
  • Others argue Anthropic “asked for this” by publicly pushing for strong AI regulation and comparing models to nuclear weapons; now those arguments are being used against them.
  • Some think rivals influenced the White House; others see it as pure Trump-era patronage, extortion, or market manipulation. Motive is widely disputed.

Model Capabilities, Safety, and Hype

  • Several users report Fable as a major step up for complex software engineering (long-horizon coding, debugging, usability review), worth the higher cost.
  • Others find it only modestly better than Opus 4.8 / GPT‑5.5, sometimes sloppier, and heavily hamstrung by aggressive guardrails.
  • Debate over danger: one side sees LLMs as serious dual-use cyber and military tech that justifies national-security controls; the other calls that “scaremongering” and notes years of doomsday rhetoric have not produced corresponding disasters.

Guardrails, Jailbreaking, and KYC

  • Consensus that “perfect jailbreak resistance” is impossible; all major models can be bypassed with clever prompts.
  • Many expect this order to push much stricter guardrails and some form of identity / citizenship verification to use top-tier models.
  • Some see this as the start of a broader push to require ID for high-capability AI (and possibly other online services).

Ecosystem & Market Impact

  • Strong concern that US closed models are now a clear supply-chain and regulatory risk for foreign users and multinationals.
  • Many predict increased investment in sovereign and open-weight models (especially from China, EU, India) and possibly relocation of AI work outside the US.
  • Others think this could cap frontier model deployment in the US, destabilize the AI datacenter boom, and hurt Anthropic’s IPO prospects—though a minority see the ban itself as powerful marketing (“so advanced the government banned it”).

User-Level Effects

  • Users report abrupt loss of Fable access mid-session, silent fallbacks, and then explicit errors; some note quotas being reset and several mention prompt refunds for recent upgrades.